A reliable SLM that outputs JSON ONLY

tools 1b

ollama run erukude/omni-json:1b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "erukude/omni-json:1b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='erukude/omni-json:1b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'erukude/omni-json:1b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 3 months ago

3 months ago

5810cbfa3aaa · 1.4GB ·

model

archllama

parameters1.24B

quantizationQ8_0

1.3GB

license

LLAMA 3.2 COMMUNITY LICENSE AGREEMENT Llama 3.2 Version Release Date: September 25, 2024 “Agreemen

7.7kB

license

**Llama 3.2** **Acceptable Use Policy** Meta is committed to promoting safe and fair use of its tool

6.0kB

adapter

archllama

parameters45.1M

quantizationF32

90MB

system

You are OmniJSON. Output exactly ONE JSON object matching this schema: keys: schema_version,intent,d

219B

params

{ "num_ctx": 2048, "temperature": 0 }

33B

template

<|start_header_id|>system<|end_header_id|> Cutting Knowledge Date: December 2023 {{ if .System }}{{

1.4kB

Readme

OmniJSON

A multi-purpose assistant fine-tuned to respond with one valid JSON object for any request (math, coding, extraction, planning, etc.). It’s optimized for deterministic structured output so apps can parse responses reliably.

Base model: Llama3.2.

Output contract

{
  "schema_version": "1.0",
  "intent": "answer|summarize|translate|extract|classify|plan|code|refuse",
  "domain": "string",
  "result": { "kind": "string", "data": {} },
  "followups": ["string"],
  "assumptions": ["string"],
  "safety": { "flags": ["string"], "notes": ["string"] },
  "confidence": 0.0
}

Training data format

[
  { "user": "...", "assistant": { "schema_version": "1.0", "...": "..." } }
]

Author

Author: Sai Teja Erukude
Role: Developer & Maintainer